Cerebellar shrinkage (especially involving the anterior superior cerebellar vermis) is among the most salient and clinically significant morbid effects of chronic hazardous alcohol consumption on brain structure. Previous magnetic resonance imaging (MRI) studies of alcohol's effect on the cerebellum relied on manual delineation, because cerebellar anatomy, with foliations narrower than standard image resolution and indistinct lateral vermis boundaries, poses particular challenges for automated segmentation algorithms. Under an SBIR Phase I grant, we (NRI) developed a novel algorithm for segmentation of the cerebellar hemispheres and vermis based on Active Appearance Modeling (AAM), which uses the prior knowledge of shape, image intensities and inter-shape relationships that human experts use to infer object boundaries in medical images. By taking advantage of recent advances in prospective motion tracking and correction that allow the acquisition of higher resolution and better quality cerebellar images, we propose to extend and optimize our cerebellar segmentation software to parcellate the cerebellum into 30 substructures, then to use our software to characterize cerebellar volumes in active alcoholics and cerebellar volume recovery during very early abstinence. We also propose to investigate how cerebrocerebellar circuits contribute to cognitive and motor impairment in alcoholism. Methods: 112 alcoholics admitted to the NIAAA for inpatient treatment protocols will be studied at treatment entry and at the end of inpatient treatment with whole brain (including cerebellar) structural imaging at 0.7 mm3 resolution corrected for movement using prospective motion tracking, the Fregly assessment of gait and balance, and cognitive assessment of abilities shown to be affected by cerebellar damage. Age and gender comparable controls will be studied with the same protocols. Expert manual delineation of cerebellar vermis and lobules (30 parcels) will be obtained from a subset of alcoholics and controls for use in developing a high-resolution statistical shape model and probabilistic atlas, which will allow us to optimize our cerebellar segmentation using the AAM approach combined with non-rigid registration to a cerebellar atlas. Our algorithm will be validated on alcoholics and controls with manually labeled test-retest images, then used to quantify cerebellar substructure volume on all participants. End of treatment and treatment entry measures will be compared in the alcoholic group to assess recovery of cerebellar structure and function during very early abstinence, and alcoholic data at the beginning and end of treatment will be compared to data from controls to estimate the magnitude of the alcohol-related damage. We will also determine whether disruption of the cerebrocerebellar circuit underlies cognitive and motor impairment in alcoholism by examining the correlational patterns between brain regions that comprise nodes of the circuit, followed by using multiple regional brain volumes to predict scores on tests of cognitive and motor functions subserved by the cerebrocerebellar circuit and known to be impaired in alcoholics.